# ComfyUI-FilePathsFromFolder Load file paths **from a folder by extension** (`*.png`, `*.mp4`, `*.txt`, etc.), then feed them into any ComfyUI node that accepts a path (image/video/audio/anything-from-path). This repo is mainly for **easy batch processing**: - Use it with your favorite loop/for/while node pack (pick items by index), or - Use this repo’s simple loop node (`File Paths For Loop Start`) to have ComfyUI run downstream nodes **once per file** (list batching). ## Installation ### Recommended: Git clone 1. Clone into `ComfyUI/custom_nodes` ```bash cd path/to/ComfyUI/custom_nodes git clone https://github.com/RockyHong/ComfyUI-FilePathsFromFolder.git ``` 2. Restart ComfyUI 3. Search nodes by name: `Load File Paths From Folder`, `Get File Path By Index`, `File Paths For Loop Start`, `File Paths For Loop Sync` ### Alternative: Manual install (ZIP) 1. Download ZIP from GitHub 2. Extract 3. Copy folder `ComfyUI-FilePathsFromFolder` into `ComfyUI/custom_nodes/` 4. Restart ComfyUI ## Integrate with external for/loop nodes (index-based) Use this when you already have an external loop/for/while node pack and you want to drive it with “files in a folder” (by extension). ### Nodes used - `LoadFilePathsFromFolder`: scan folder → get `filepaths` / `filenames` / `count` - `GetPathByIndex`: choose one item → get `filepath` / `filename` (UI: `Get File Path By Index`) ### Example workflow (external loop pack) ![Option A — External Loop Method](screenshots/Option%20A%20(External%20Loop%20Method).jpg) 1. `LoadFilePathsFromFolder` (set folder path + extension) 2. `count` → your loop node “iterations” / “max” (names vary by pack) 3. Loop `index` → `GetPathByIndex.index` 4. `LoadFilePathsFromFolder.filepaths` + `.filenames` → `GetPathByIndex` 5. `GetPathByIndex.filepath` → your loader node (Load Image / Load Video / etc.) 6. Do your processing + save/export inside your external loop pack ## Built-in simple for-loop node (list batching) Use this when you want the simplest setup: `File Paths For Loop Start` returns lists, and ComfyUI runs downstream nodes **once per file** via list batching. ### Nodes used - `File Paths For Loop Start`: generates the per-file list outputs (`current_filepath`, `current_index`, etc.) - `File Paths For Loop End`: loop terminator / completion signal ### Example workflow (internal loop: once per file) ![Option B — Internal Loop Node Method](screenshots/Option%20B%20(Interal%20Loop%20Node%20Method).jpg) 1. `File Paths For Loop Start` (set folder path + extension) 2. `current_filepath` → your loader node (e.g. Load Image) 3. Your processing nodes 4. (Optional) outputs you want to **guarantee run** (e.g. save/write/export nodes) → `FilePathsEnsureExecution.passthrough_1` … `passthrough_5` 5. `current_index` → `File Paths For Loop End.current_index` 6. (Optional) `FilePathsEnsureExecution.passthrough` → `File Paths For Loop End.passthrough` (this “anchors” the route so it executes) ### Memory note `File Paths For Loop Start` uses **ComfyUI list batching** (`OUTPUT_IS_LIST`): ComfyUI may run the workflow **one node at a time for all files**, rather than finishing file 0 completely before starting file 1. - **What you'll see**: it may run like **"load all → process all → save all"**, not "load → process → save" per file. - **When this is good**: if you want to run many files with the **same model/settings**, batching can help keep the model "warm" and avoid offload/reload each file. - **Why memory can climb**: more intermediate results may stay in RAM/VRAM until later nodes run. - **Solution for mid-batch cleanup**: Use `File Paths For Loop Sync Trigger` + `File Paths For Loop Sync Resume` to run cleanup **once** between batch stages (see [Batch Synchronization](#batch-sync) below). - **Need true one-file-at-a-time?** Use an external loop/for/while node pack that runs each file start→finish before moving on. ### (Optional) Ensure execution for save/write/export routes Some ComfyUI workflows can “skip” routes that don’t affect the final output the graph thinks it needs. If you have save/write/export nodes that you **must** run, use `File Paths For Loop Ensure Execution` as an anchor. ![File Paths For Loop Ensure Execution](screenshots/File%20Paths%20Loop%20Ensure%20Execution.jpg) - Connect the nodes you want to guarantee run into `passthrough_1` … `passthrough_5` - Connect `passthrough` output to `File Paths For Loop End.passthrough` ## Batch synchronization (run once between stages) Use this when you need to run something **exactly once between two batched stages** (e.g. LLM batch → clear VRAM once → T2I batch), without switching to an external per-file loop. ![File Paths For Loop Sync](screenshots/File%20Paths%20Loop%20Sync.jpg) - **`File Paths For Loop Sync Trigger`**: put this right after a batched output. It provides: - **`anything`**: the original batch (passes through unchanged) - **`trigger`**: a single (non-batched) value you can feed into **any “run once”** node - **`File Paths For Loop Sync Resume`**: a simple wait point. It: - Takes the batch on **`anything`** (from Trigger) - Takes a completion signal on **`wait_for`** (from your one-time node) - Outputs the same batch so downstream nodes continue as a batch **How to wire it (in words):** connect your batched data into `Sync Trigger` → connect `Sync Trigger.anything` into `Sync Resume.anything` → connect `Sync Trigger.trigger` into your one-time node → connect that one-time node’s output into `Sync Resume.wait_for` → continue the workflow from `Sync Resume.anything`. **What you can run once:** clear RAM/VRAM, save/checkpoint, switch models, send notifications, write logs, etc. ## License MIT — see `LICENSE`.